Modeling the Nonlinear Dynamics and Nonstationarities Underlying Spike Train Transformations in the Brain

• Activity Number: 114
• Type: Topic Contributed
• Date/Time: Monday, August 1, 2011 : 8:30 AM to 10:20 AM
• Sponsor: Section on Nonparametric Statistics
• Author(s): Dong Song*+ and Theodore W. Berger
• Companies: University of Southern California and University of Southern California
• Keywords: point processes ; GLM ; hippocampus ; spike ; prosthesis ; Volterra model

Abstract

Brain represents and processes information with spikes. To understand the biological basis of brain functions, it is essential to model the spike train transformations performed by brain regions. Such a model can also be used as a computational basis for developing cortical prostheses that can restore the lost cognitive function by bypassing the damaged brain regions. We formulate a three-stage strategy for such a modeling goal. First, we formulated a multiple-input, multiple-output physiologically plausible model for representing the nonlinear dynamics underlying spike train transformations. This model is equivalent to a cascade of a Volterra model and a generalized linear model. The model has been successfully applied to the hippocampal CA3-CA1 during learned behaviors. Secondly, we extend the model to nonstationary cases using a point-process adaptive filter technique. The resulting time-varying model captures how the MIMO nonlinear dynamics evolve with time when the animal is learning. Lastly, we seek to identify the learning rule that explains how the nonstationarity is formed as a consequence of the input-output flow that the brain region has experienced during learning.